Health workers’ strikes in low-income countries: the available evidence

Abstract Objective To analyse the characteristics, frequency, drivers, outcomes and stakeholders of health workers’ strikes in low-income countries. Methods We reviewed the published and grey literature from online sources for the years 2009 to 2018. We used four search strategies: (i) exploration of main health and social sciences databases; (ii) use of specialized websites on human resources for health and development; (iii) customized Google search; and (iv) consultation with experts to validate findings. To analyse individual strike episodes, pre-existing conditions and influencing actors, we developed a conceptual framework from the literature. Results We identified 116 records reporting on 70 unique health workers’ strikes in 23 low-income countries during the period, accounting for 875 days of strike. Year 2018 had the highest number of events (17), corresponding to 170 work days lost. Strikes involving more than one professional category was the frequent strike modality (32 events), followed by strikes by physicians only (22 events). The most commonly reported cause was complaints about remuneration (63 events), followed by protest against the sector’s governance or policies (25 events) and safety of working conditions (10 events). Positive resolution was achieved more often when collective bargaining institutions and higher levels of government were involved in the negotiations. Conclusion In low-income countries, some common features appear to exist in health sector strikes’ occurrence and actors involved in such events. Future research should focus on both individual events and regional patterns, to form an evidence base for mechanisms to prevent and resolve strikes.


Introduction
Workers' strikes or industrial action are identified as "…the collective withholding of labour/services by a category of professionals, for the purpose of extracting concessions or benefits, typically for the economic benefits of the strikers." 1 The right to strike is widely considered a civil right and is often part of a country's legal system. 2 However, for some professional groups, including health workers, strikes might have implications beyond the involved parties. Health workers' strikes has been purported as putting patients at risk of serious harm and potentially contradict health workers' duties to care for their patients 1,3 and evidence from highincome settings shows that nurses' strikes can affect hospital mortality. 4 Although doctors' strikes in high-income settings may not necessarily increase patients' mortality in the short term, 5 they can severely disrupt the provision of health-care services, with significant political, organizational and financial implications. 6 In middle-income countries, some evidence shows that physician strikes can lead to a decrease in clinical activities and increase in mortality. 7,8 Some have argued that such strikes would be justified if directed towards improving workers' conditions and their ability to care for future patients 9 and that doctors' strikes may be morally acceptable if proportionate and properly communicated. 10 Health workers' strikes are of growing concerns to the international health community and organizations aiming to ensure health and access to health services for all. Health workers play a central role in achieving universal health coverage (UHC), and interruptions in health services not only hold implications for UHC, but suggest unresolved labour and governance issues in health sectors, particularly in some lower-income settings with poor governance and regulations.
Determinants of health workers' strikes can be diverse. For example, gross domestic product (GDP) growth and widening wage differentials across professions have been associated with industrial action in the private sector in the United States of America. 11 Compensation differentials appear to be a central determinant of public sector strikes, as wage disputes are informed by comparisons with private sector salaries, or with public sector salaries of or similar-level professions. 12 Economic theorists suggest that the likelihood of strike action increases when a country's general economic conditions improve and unemployment rates are low, 11 because renegotiation on how to share the society's increased wealth among its members is needed. Improving the profession's social and political position within the society can be another motivation for strikes. 13 Currently, the main mechanisms for strike resolution in most high-income countries are outright prohibition of strikes for certain public sectors, formal systems for impasse procedures and forms of binding arbitration between the negotiating parties. 12,14 In the public sector, formal systems of impasse procedures, such as conventional, binding and final offer arbitration, are often in place, where governments assist employers and unions to help resolve disputes. 12 Medical associations and councils play a critical role in not only triggering, but also mediating strikes because of their role of unions and self-regulatory bodies. 15 For example, in Mozambique and the United Republic of Tanzania, medical associations were formed to negotiate remuneration grievances from junior doctors, in contrast to the official medical councils Health workers' strikes in low-income countries Giuliano Russo et al. representing the interests of more senior cadres (Arroz J, Associação Médica de Moçambique, unpublished data, 2014). 16 In low-income countries, data on health workers' strikes are scarce and the implications of such strikes are potentially wide reaching when health systems are fragile. This paper analyses the evidence on health sector strikes in low-income countries in the last decade with the aim of understanding their characteristics and drivers and providing a baseline for future research.

Search strategy
We adapted a published method 17 to systematically search the grey literature, focusing primarily on online resources given that health workers' strike episodes are more likely to be reported by the media than in academic publications. For this work, we used the Luxembourg's definition of grey literature, covering documentation produced by all levels of government, academics, business and industry in print and electronic formats, produced by non-commercial publishers. 18 First, we searched general health and social science databases (PubMed®, Scopus, EconLit and Web of Science) between July and August 2018 using the following search terms: "Country name" AND "strike" OR "industrial action" AND "physicians" OR "doctors" OR "nurses" OR "pharmacists" OR "dentists" OR "midwives" OR "health worker" OR "hospitals. " We then searched specialist development databases (ReliefWeb, World Health Organization's (WHO's) Index Medicus and Global Nonviolent Action Database) and dedicated websites on labour and human resources for health (Public Services International, ILO, WHO and the World Bank) between July and September 2018 using combinations of the above search terms.
We then conducted customized Google searches (Box 1) of media reports in English and if we did not identify any record in English, we searched in one of the official languages of the countries reviewed (Spanish, French and Portuguese). The search covered international websites of BBC News Africa, Al Jazeera, Fox News, France 24 Observers, Thomson Reuters Foundation News, Reuters and the local news networks Mail & Guardian, AllAfrica, MedAfrica Times, Medical Xpress, Guinee Matin, Caribbean Life News, Africanews, as well as on national Medical Associations news databases, using combinations of the above search words and their correspondent in the local language. For identified records in local languages spoken by at least a million people in one or more of the low-income countries (Amharic, Pashtun and Swahili), we used Google Translator. The customized searches were updated in March 2019 (number of Google hits are available from the corresponding author).
Finally, for countries which we did not obtain any information on health workers' strikes, one of the authors consulted nine local and international health sector experts in June and August 2018. To improve the comprehensiveness of this strategy, we chose content experts to reflect a diversity of disciplines and geographical areas relevant for the strike events (list of experts available from the corresponding author).

Inclusion and exclusion criteria
We used the World Bank's 2017 classification to identify the 31 low-income countries for our search. 19 Titles were included if they reported on "a temporary work stoppage effected by one or more groups of workers with a view of enforcing or resisting demands or expressing grievances or supporting other workers in their demands or grievances. " 20 We included reports of events that took place between January 2009 and December 2018.
Information on strikes based on only social media reports were excluded due to the inability to triangulate and validate the information. Titles on alternative forms of industrial action were not included, such as go-slow strikes, threats to strike or silent protest marches. Given the inclusion of macroeconomic factors and a need to ensure consistency across health-care settings, only nation-wide or province-wide strikes were considered. Strike episodes that were highly localized (e.g. in a single hospital) and reports on threats of strike action were excluded.

Framework
To guide the data extraction and the analysis of the individual strike events identified through our search, we developed a framework that summarizes the concepts from the economic, political economy and health system research literature on health sector strikes, and helps to understand the linkages between pre-existing conditions, relevant influencing actors and their interaction (Fig. 1).
The framework highlights the need to consider the micro as well as macro dimensions of health sector strikes. For example, pre-existing economic and legal conditions, including economic growth, wage and unemployment levels, or the existence of mechanisms for resolution of disputes, are associated with strike onsets. Actors, such as unions, government, parties and professional associations, play a role in driving and resolving the disputes.
The approach to develop this framework was in line with the High-Level Commission on Health Employment and Economic Growth, 21 which highlights the interaction between health workers, the health sector and the macroeconomic context of a country. The framework also draws from the method used by the Organisation for Economic Cooperation and Development (OECD)

Box 1. Customized Google searches for health worker's strikes in low-income countries
We searched information for each of the 31 low-income countries using the search string in English: "[country name] strikes physician" OR "doctor" OR "nurse" OR "pharmacist" OR "dentist" OR "midwife" OR "health worker" OR "hospital. " For countries with French as official languages, we used a search string in French if we did not identify any records using the English search string: "[Nom du pays] grève médecin" OR "infirmier" OR "pharmacien" OR "dentiste" OR "sages-femme" OR "agent de santé" OR "hôpital" For countries with Spanish as official languages, we used a search string in Spanish if we did not identify any records using the English search string: "[Nombre del país] and huelga" OR "medico" or "enfermera/o" OR "farmaceutico" OR "dentista" OR "comadrona" OR "and trabajadores del sector salud" OR "hospital" For countries with Portuguese as official languages, we used a search string in Portuguese if we did not identify any records using the English search string: "[Nome do país] and greve" OR "doutor" OR "enfermeiras" OR "farmacêutico" OR "dentista" OR "parterira" OR "trabalhadores do sector saúde" OR "hospital" Research Health workers' strikes in low-income countries Giuliano Russo et al.
and International Labour Organization (ILO) to collect data on strikes and collective bargaining systems, and to analyse the impact of such events on labour markets in high-income countries. 22

Data extraction
We created two Excel spreadsheets with the pivot table feature (Microsoft, Redmond, United States of America) to store and analyse the information from records included in the study.
All identified records were screened by two authors. From records meeting the inclusion criteria, they extracted data on the length of strike episodes, main actors involved, relevant strike features and resolutions of events, following the conceptual framework. For validation, we triangulated the information on unique strike events identified from the eligible records. We regarded international news sources as more trustworthy than national and local ones. When two sources gave conflicting accounts of causes and resolution, we sought for a further source of information for clarification.
We extracted GDP per capita (in United States dollars) and GDP growth data for the years 2009 to 2016 from the World Development Indicators database. 19 For the years 2017 and 2018, we obtained current GDP per capita and country GDP growth from the International Monetary Fund. 23 We retrieved information on unemployment rates from the ILOSTAT database 24 for the years 2009 to 2017. Unemployment rates for 2018 were either incomplete or not available by March 2019, we therefore assumed they were unchanged from 2017.
We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines 25 to report on the review.

Results
The review of the general health and social science databases yielded an initial 34 titles from the published literature, three of which we retained after screening. Searching specialist databases resulted in 91 potentially relevant titles, of which five reports on individual health workers' strikes were eligible for inclusion. We identified an initial 676 records, of which 109 met the inclusion criteria after elimination of duplicates (available from corresponding author) when doing the customized Google searches. In total, 116 reports covering 70 unique strike episodes in low-income countries met our inclusion criteria. Of the reports identified, most (103) were online media reports, five human resources for health reports from ReliefWeb and The World Bank databases and two academic publications (Fig. 2). 16,26 We identified strike episodes across 23 low-income countries between 2009 and 2018 (Table 1; available at: http://www.who.int/bulletin/ volumes/96/7/18-225755). Eight lowincome countries had no report of health workers' strikes during this period (Afghanistan, Central Africa Republic, Eritrea, Ethiopia, Democratic People's Republic of Korea, Guinea, Myanmar and Rwanda). The experts identified six initial records from these countries, however, none of these records were eligible.
All included health workers' strikes were suspension of service provision, with only emergency services guaranteed in hospitals' emergency and resuscitation departments.

Frequency and duration
The median number of strike events was six per year, however, the data collected show an irregular pattern of episodes over the decade, with most strikes (49 events) recorded in the last five years. The years 2014 and 2018 registered the highest number of episodes, 10 and 17 events, respectively (Fig. 3). The year 2018 had the highest number of total work days lost (170), while Niger recorded the largest number of reported strikes (seven events), followed by Sierra Leone and Zimbabwe (six events; Table 1).
From the records reporting on number of days of health workers' strikes, we calculated that a total of 875 working days were lost between 2009 and 2018, with a median number of 77.5 working days lost per year. That is, on every third working day on average, there was a strike taking place in the health sector in a low-income country during this period. Strike episodes lasted an average of 12.5 days, although some strikes protracted for months, such as the general health sector strikes in Haiti in 2016. Some strikes were recurring for months or years (as in Burkina Faso between 2012 and 2018, in Niger between 2011 and 2017 or Zimbabwe between 2014 and 2018)

Economic and political conditions
Complaints about inadequate remuneration and delayed payments, were the most common causal factor cited (90% of events; 63/70), followed by protest against the slow implementation of a previously reached agreement, or against the health sector's governance and policies (36%; 25/70). Complaints about working conditions and security issues were mentioned in 14% (10/70) of the events.
Strike episodes were reported during years of weak as well as strong GDP growth, with a median growth of 4.51% (standard deviation, SD: 1.96) and an unemployment rate of 5.12% (SD: 2.80) in the affected countries (Table 2). Although strike episodes appeared to be more frequent in more recent years, no specific variable was identified for this pattern. We found little quantitative information on salary differentials between the public and private sector, but in several cases salary levels for other public servants were reported to be a reference in the negotiations (such as for physicians and senior levels of the judiciary for Mozambique in 2013, and for junior and specialist doctors in Niger 2017).

Actors involved
We identified 62 reports containing information about stakeholder involvement, including professional trade unions (general and health sector specific), medical and clinical associations and government authorities in charge of negotiations (health ministry, finance ministry, President, Prime Minister or Cabinet). Striking parties were represented by professional associations, and by diverse government institutions, such as the health ministry, Presidency, Prime Minister Office and the finance ministry. Health professional councils and associations, rather than general trade unions, were involved in all the strikes identified.
Industrial action involving more than one professional category was the most common strike modality (46%; 32/70 of strike events reported), followed by strikes by physicians only (31%; 22/70 of strike events reported). Only in Zimbabwe in 2018 we found reports of nurses striking independently from other health professionals.
Reports of violent confrontation with the government were found in four cases. No explicit mention of specific mechanisms of dispute resolution was found in the reports.
Resolution was more frequently reached when other ministries (finance or public administration ministry) or higher levels of decision-making (such as Prime Minister or President) were involved, rather than the health ministry alone. According to the reports, external international actors were rarely involved in the negotiations, with the notable exception of human rights nongovern-mental organizations (NGOs) in the United Republic of Tanzania in 2012 and Chad in 2018, and the World Bank's intervention in Guinea Bissau's health and education workers' strike. 28

Discussion
This study analyses health workers' strikes in low-income countries and links the phenomenon to a theoretical framework. Future studies will be able to build on this baseline study and use it for monitoring trends. As we mostly extracted information from online 34   Research Health workers' strikes in low-income countries Giuliano Russo et al.

Fig. 2. Flowchart of the included records on health workers' strikes in low-income countries
media and press reports, the study provides some unknown level of comprehensiveness. Volumes of internet users and reports from low-income settings have evolved unevenly in recent years and therefore our searches might have missed information from countries with lower access to internet services. Although our findings are not fully comparable to the OECD data on work days lost to strikes per thousand workers in high-income labour markets, 22 our results show that in low-income countries health workers' strikes have become more frequent in recent years. However, the consequence for the patients, due to the disruption of health-care provision for a substantial number of days over the decade, is unknown. Understanding and monitoring heath workers' strikes is therefore important, as such events could slow down the progress of achieving UHC.
We were not able to find reports of health workers' strikes for eight low-income countries during the years 2009 to 2018. This could be due to several factors. First, information may not have been readily available on the internet for these countries. Second, a substantial portion of health workers have been employed by international NGOs in these countries, making public sector industrial action less noticeable. Third, public health workers may also be engaged in private provision of services, therefore reducing the impetus of strikes. 29 Finally, in some countries strikes in the health sector are simply not permitted. 30,31 Although wage demands were central to most of the strike events reviewed, macroeconomic conditions, such as GDP growth, unemployment and absolute salary levels, did not appear to be key triggers. Relative pay gaps between junior and senior cadres or with other professions were mentioned as a more frequent source of recrimination.. 32,33 Our data were not sufficient to allow the identification of specific polictical economy factors for the strike episodes; however, our results do suggest that professional associations, government departments, health sector and labour market governance, all contribute in reaching positive resolutions. In physician' cases, as senior doctors have traditionally been well-connected with the government, they have had more effective means of influencing governments and to protect their economic interests. 34 Therefore, strikes may arise from the failure of the medical associations to represent more junior doctors or general practitioners. 13 To advance the understanding of health workers' strikes, the political economy aspects of individual strikes and the implications of political actors contributing to positive resolutions need to be considered. Furthermore, investing in the development of collective bargaining systems may help reduce the scope for strikes. 22 Our results suggest that health sector strikes are context-specific, but also share some commonalities. An appropriate research agenda should therefore encompass both case-studies of individual events and more general region-wide studies looking into wider patterns of causality. Different disciplines, including economics, sociology and political science, have so far offered isolated angles and interpretations of health sector strikes. We believe that a more integrated multidisciplinary approach would be more suitable for untangling the factors of such strikes and provide an evidence base for positive resolution of such conflicts. Better understanding of strike triggers and pathways to resolution could improve the sector's governance, patients' access to services, and ultimately, the achievement of UHC goals. ■